Management: Adjusting to the Newest Artificial Team Member

(1)

Management: Adjusting to the Newest Artificial Team Member

How does AI implementation affect the role, competencies, and status of management?

Name: Max Goldman

Student number: 13447068 Date: 17-08-2022

Version: Final

Qualification: MSc. Business Administration – Digital Business Track University: University of Amsterdam (UvA)

EBEC Number: EC 20220606020643 Supervisor: Beauregard Berton Second assessor: Rick Hollen

(2)

Statement of Originality

This document is written by Student Max Goldman who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1 Introduction

“Artificial Intelligence, deep learning, machine learning — whatever you’re doing if you don’t understand it — learn it. Because otherwise you’re going to be a dinosaur within 3

years.” (Cuban, 2017)

This quote dates back five years at the moment of writing and it looks like we’re not there yet, but to what extent do we need to understand Artificial Intelligence (AI) within today’s organization? With AI being prognosed to contribute in the deliverance of all products and services (Chui et al., 2018) and other researchers going even further with believing that technological advancements in AI will cause a technological revolution that will transform humanity (Brock & von Wangenheim, 2019). The attention for AI has been increasing and this has driven rapid progress in recent years. Processes like recruitment selection, spotting melanoma, self-driving cars, AI-generated artwork, are all being automated with AI

implementations. Since the computational power has increased, the adoption of AI is

contributing to overall firm performance, for instance lowering costs if part of the recruitment process is automated or increasing revenue if the marketing is personalized through AI

(Mikalef & Gupta, 2021).

Business is increasingly focused and therefor investing in the developments of AI, and this is especially visible in the insurance industry, which has implemented AI in processes such as fraud detection, claims processing and personalized consumer targeting (Soberanis, 2021). According to Plastino & Purdy (2018), compared to other sectors, the implementations of AI in the financial services sector will have most impact at multiple levels of the organization and society. The impact is not always perceived as positive with employees who are required to cooperate with AI in their day to day activities whilst they are afraid of losing their jobs to the same system (Makarius et al., 2020). Also perceived as

(5)

negative is the impact on society as biases in the algorithms are placing certain groups systematically at a disadvantage (Ferrer et al., 2021).

Raisch & Krakowski (2021), however indicate that much is still undiscovered about the use of AI in organizations and stress the importance for management scholars to be more involved. This is underpinned by Reim et al. (2020) as they elaborate: “Managers are left with little support from academia when aiming to implement AI in their firm’s operations which leads to an increased risk of project failure and unwanted results”. With organizations investing in AI but failing to achieve their goals, there is a need for a better understanding of the required human skills. This research is focused on management that is affected by these implementations as, for instance, both the ethical and the employee perspective needs to be considered in their management role. Furthermore, a recent survey of MIT Sloan

Management Review indicates, that the lack of leadership support is ranked among the top barriers for successful AI implementation (Ransbotham et al., 2018). This is also elaborated upon by Mikalef & Gupta (2021), as they point out that a manager needs to understand and commit to the change, acquire the ability to foresee potential applications of AI and then needs to handle the transition to AI enabled processes. It calls for different competencies as Kolbjørnsrud et al. (2016) elaborates that management needs to be able to initiate and plan AI implementations. This is a real challenge for current management, as a recent study reveals that 37% of managers do not understand how AI technology works (Davenport & Ronanki, 2018). In addition, as the number of AI implementations increase, the role of technical employees in the organization will become more dominant, while the jobs on the operational side will be replaced (Kiron, 2017). Managing the relationship on both sides will be crucial for successful adoption of AI implementations (Kiron, 2017; Kolbjørnsrud et al., 2016;

Mikalef & Gupta, 2021). Looking at the power constructs of Finkelstein (1992), the status of management and especially expert power will be under pressure as the intellectual and

(6)

cognitive proportions of technology have no boundary and will in time (if not already) outperform the cognitive capabilities of humans (Frey & Osborne, 2017; Haenlein & Kaplan, 2019). With these developments, management is under pressure to adjust accordingly to maintain their position and to enable successful adoption of AI within the organization (Andrews, 2017). This stresses the importance of this research, as it explores the changes needed in management due to AI, and also for AI implementation and adoption in

organizations.

Management literature has provided many theories over the last decades, however to stay in line with previous AI and management research, the purpose of this research is to explore the changes in the role, competencies and status of management due to AI

implementation. Raisch & Krakowski (2021), also indicate in their research that one of the gaps in current literature is the effect of AI implementation on the role, competencies, and status of management. As the prior research has not explored the specific changes of management due to AI implementation, this research has taken an inductive approach to develop new theory. By conducting semi-structured interviews with managers in the insurance industry and by applying the Gioia methodology, this research contributes by assisting management in making the right adjustments and by helping organizations with the development of the right management AI capabilities (Mikalef et al., 2021) as well as building upon existing theory on the role, competencies, and status of management as organizations are becoming increasingly AI driven. The insights of the changes in their role, competencies, and status, will allow this research to answer the following Research Question (RQ):

“How does AI implementation affect the role, competencies, and status of management?”

(7)

2 Literature review

Although little is known about the changes in management role, competencies, and status due to AI implementation, this chapter elaborates on the existing theory of AI, the related

management theories and the current implications of AI implementations in the insurance industry. This provides a base for the findings chapter and the creation of new conceptual models.

2.1 Artificial Intelligence

AI is researched in different settings ranging from creative industries to strategic decisions to the actual science behind it. Haenlein & Kaplan (2019) look at the different angles and state that AI is commonly defined as: “a system’s ability to interpret external data correctly, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation”. The interpretation of external data and applying this knowledge to achieve specific goals, is something closely related to human decision making (Krakowski et al., 2022). An important distinction between the two is however that humans rely on data processing shortcuts, which sometimes leads to an error or biased decision but also offers more flexibility. The function of AI in this research could be both as complementation as well as the substitution of human, or as Raisch & Krakowski (2021) refer to it as AI augmentation or automation. The following sub-chapter will elaborate on the related AI capabilities and the impact on current organizations.

2.1.1 AI capabilities

Organizations have used AI for routine tasks in operations for several decades (Raisch &

Krakowski, 2021), and pioneers in the 1950s already predicted that AI would even affect managerial tasks (Newell et al., 1959; Newell & Simon, 1956). As the progress of technology was slow, the expectations lowered and any impact on management level was seen as too

(8)

comprehensive (Cariani, 2010). Research on the implications for AI were therefore partly disregarded in the years that followed (Kellogg et al., 2020).

As computational power and the amount of data increases, the function of AI now ranges from increasing efficiency and accuracy on the process level to strategic decision making on the organizational level (Davenport et al., 2020; Davenport & Kirby, 2016; Ghosh et al., 2019). This potential has driven many industries and organizations to incorporate AI (Daugherty & Wilson, 2018). The ability to self-learn and scale-up leads to new possibilities in achieving a competitive advantage (Frey & Osborne, 2017; Plastino & Purdy, 2018). As such, AI is often assisting decision making with valuable insights, derived from big data sets that are compressed into manageable scale (O’Leary, 2013). To give an example, Netflix uses an algorithm to give personalized recommendations to its customers, which creates higher customer loyalty or to decide on which movie plots to pursue (Gomez-Uribe & Hunt, 2015;

Westcott Grant, 2018). Another example is used for Human Resource Management (HRM), where most decision within the selection process are based on AI input, which creates a possibility for increased scalability and accuracy (Tschang & Almirall, 2021).

AI is integrated more and more in organizations as AI is already involved in almost all products of the Internet of Things (IoT). Some business innovators go even further and believe that AI will be incorporated in almost every service (Chui et al., 2018; Raisch &

Krakowski, 2021). Most AI implementations are functions related to Natural Language Processing (NLP), robotics automation, picture analysis and machine learning and treated as a disruptive innovation (Burrell, 2016; Haenlein & Kaplan, 2019). It is influencing the rules of all industries and, as with all disruptive innovations, there are early and late adaptors (Brock & von Wangenheim, 2019; Reim et al., 2020). The late adaptors in this sense are not able to exploit the AI impact which is especially seen in the failure to attract talent and to

(9)

develop internal AI capabilities, whereas the front runners are able to capture the best opportunities (Bughin et al., 2018).

The term AI capabilities is still relatively new, as it followed, the before mentioned, increase of AI implementations in business (Mikalef & Gupta, 2021). The concept is

grounded on more general Information System (IS) research which captures the capability of the organization to get effective new technology deployment, in contrast to only adopting the technology infrastructure (Conboy, Mikalef, et al., 2020; Handali et al., 2020). The IT

capabilities, refer to the ability to apply human, organizational and technology resources, in order to add value (Kim et al., 2011; Liu et al., 2013). This results in a better estimation of possible value creation as it involves the intangible parts of enabling technological

implementations (Conboy, Dennehy, et al., 2020). The AI capabilities follows this concept, as it also looks at the organizational and technical capacity to leverage AI implementation. It is therefore defined as follows; “AI capability comprises of complementary AI-related tangible, human, and intangible resources.” (Mikalef et al., 2021).

Tangible resources in this context, include the data, the technical infrastructure, and the processing power (Aboelmaged, 2014; Desouza et al., 2020; Duan et al., 2019).

Intangible resources on the other hand, include the coordination across departments, the capacity available to carry out the organizational change and the readiness of the organization to do projects that are high-risk high-return (Davenport & Ronanki, 2018). Finally, there is human resources which includes the balance of both managerial and technical skills.

Managerial skills are required to identify the domain knowledge needed for AI

implementations and envisioning the right areas within the organization, while technical skills are needed for data handling and AI techniques.

(10)

2.1.2 AI adoption

Organizations often struggle to successfully adopt AI because they fail to focus also on developing internal human skills (Andrews, 2017). Acquiring the needed skills is difficult as the current workforce has a shortage in supply of technological skilled employees (Danyluk

& Buck, 2019). However this is not the main reason for not reaching organizational goals, as a recent survey of MIT Sloan management review indicates that the lack of leadership support is ranked as the top barrier for successful implementation of AI (Ransbotham et al., 2018). Mikalef & Gupta (2021) indicate that management needs to understand and commit to the change, acquire the ability to foresee potential applications of AI, and needs to develop an understanding on handling the transition to AI enabled processes. This is a real challenge for current management, as a recent study reveals that 37% of managers do not understand how AI technology works (Davenport & Ronanki, 2018). Current management will need to initiate and plan AI implementations, which requires different competencies (Kolbjørnsrud et al., 2016). In addition, as the number of AI implementations increase, the role of technical employees in the organization will become more dominant, while the jobs on the operational side will be replaced (Kiron, 2017). Managing the relationship on both sides will be crucial for successful adoption of AI implementations (Kiron, 2017; Kolbjørnsrud et al., 2016).

2.2 Management

As organizations try to stay ahead of competition, AI and data are increasingly important and legal as well as ethical boundaries keep changing (Kolbjørnsrud et al., 2016). This affects, as mentioned above, what is expected from management. Larson & DeChurch (2020) take a broader look and identify four different influences of technology on management within the research enterprise over the last 50 years; digital technology as team context, digital

technology as sociometrical team practices, digital technology as team creation medium and digital technology as a teammate. All the developments create different needs from

(11)

management. An example would be that with technology as team context, a challenge to reach team core needs was created as virtual collaboration was introduced, resulting in several team members showing leadership behavior also referred to as shared leadership (Charlier et al., 2016; Contractor et al., 2012).

The first three technology perspectives consider technology affecting teamwork, but technology as a teammate does not imply constraining or augmentation of teamwork, rather it views technology as a co-equal teammate (Larson & DeChurch, 2020). This form of working alongside machines has been around for several decades with Alen Turing first exploring the possibility of technology as code cracker (Cawthorne, 2014). With AI becoming more embedded in all parts of the organization, management needs to foster the core needs of their team, that now includes AI, in order to improve firm performance (Larson & DeChurch, 2020; Mikalef & Gupta, 2021). Furthermore, the team has to have a shared understanding of possible limitations of AI and when an intervention is needed (Larson & DeChurch, 2020).

AI is affecting the entire organization and management needs to evolve and adjust (Kolbjørnsrud et al., 2016).

2.2.1 Management theory

As the RQ already indicates, this research is focused on the role, competencies and status of management being affected by AI implementation. Management theory is of course more diverse than just these three, with the accepted leadership theories like great man, trait, contingency, transactional and transformational, to name a few (Bolden et al., 2003). These all have interesting angles on management however, this research will focus on the changes in competencies, role and status which are discussed in the following parts.

(12)

2.2.2 Management Competencies

This research looks at competencies as defined by Moore et al. (2002) as “the recognition of the level of competence for a professional deriving from their possessing a number of

relevant attributes such as knowledge, skill and attitudes”.

As mentioned in AI capabilities, there is a need for a different manager in terms of knowledge, skills and attitudes. The digital age in general has driven the manager to several adjustments; as an example the transition to virtual meetings during the Covid-19 pandemic changing the skills and attitude needed to have productive meetings with a team (Neumeyer

& Liu, 2021). AI capabilities theory mentions the importance of managements’ AI

knowledge, enabling them to foresee future opportunities in their domain (Mikalef & Gupta, 2021). Looking at management capabilities, the AI implementation makes knowledge generalizable which might be indicating that profession knowledge is no longer as important as it used to be.

As organizations are implementing AI, decision making of managers is increasingly shared with AI (Haesevoets et al., 2021). Most managers believe that the majority vote of the decisions should remain human, with a preference for 70% human and 30% AI based

(Haesevoets et al., 2021). However as elaborated some of the managerial decisions are already taken over by machines and especially the large technology organizations such as, Amazon, Google and Netflix, have automated many decision-making processes on their management level (Schrage, 2017). Haesevoets et al. (2021), emphasize in this that it is important to think from an AI perspective, meaning that management should not look on how humans fit into the process of AI decision making, they should look how AI can contribute to the team. This highlights that management needs to be capable of foreseeing the added value of AI.

(13)

Data is, as mentioned before, is one of the elements that drives the AI capability of an organization (Haenlein & Kaplan, 2019; Mikalef et al., 2021). The governance of this data, especially the confidential data which is used in the insurance industry, is crucial to manage (Reim et al., 2020). This competence is most likely needed from a management perspective but will also include training and development of the team (Reim et al., 2020). Experts in the field of AI are scarce and internal development might be the only option for many

organizations, stressing the importance of the managers’ competence to enable their employees to develop (Petropoulos, 2018).

2.2.3 Leadership role

Mintzberg (1989), concluded that a management job could be described with ten roles or as he describes it as “organized sets of behavior”. As scholars build upon his theory a generalization was made by Morgeson et al. (2010), who identified two dimensions:

leadership functions and forms. Derived from these dimensions, Morgeson et al. (2010) define the leadership role as the ability to achieve team effectiveness.

Leadership functions, are focused on enabling teams to reach their core needs and to foster team effectiveness (Morgeson et al., 2010). The core needs of the team are to develop affective emergent states, to develop cognitive emergent states and to enact behavioral integration processes (Morgeson et al., 2010). Affective states are the degree of satisfaction and the identification with the organization, which is under pressure as AI might cause job replacement (Morgeson et al., 2010). Cognitive states are to learn and adapt to the team over time, which is probably influenced because AI is seen as a part of the team (Haesevoets et al., 2021; Morgeson et al., 2010). Behavioral integration is the degree of helping another team member or to engage in prosocial behavior for the larger organization (Morgeson et al., 2010). Leadership forms, are focused on how leadership is carried out (e.g. hierarchical, shared, distributed and rotated leadership), where theory emphasizes that collective forms of

(14)

leadership reach better results of team effectiveness (Contractor et al., 2012). These two dimensions lead to team leadership, that can be seen as the actions of one or multiple team members to secure that team needs are matched.

The call for a different leadership role increased as virtual teams, online communities, crowd sourcing, human-AI and robot teams were introduced (Larson & DeChurch, 2020).

The kind of manager needed has shifted and this only increased as the pandemic of Covid-19 disrupted organizations (Coombs, 2020). As AI is seen as a disruptive innovation, as

mentioned earlier, it is expected that a manager needs to take on different leadership roles within the organization.

With reference to the AI capabilities chapter, the leadership role after AI

implementation needs to adjust to employees on the technical side having a more dominant role in the organization and the operational side showing resistance due to fear of job loss.

Resistance is regarded as behavior that is not according the effort of the change leader (Bartunek, 1993). In management science there are typically two: “Prescriptions are given to those with formal, legitimate power and influence over the change (change agents) for how to avoid and overcome the resistance of those without it (resisters)” (Van Dijk & Van Dick, 2009, p. 144). Although in this definition managers could be among the resisters, the AI capabilities point out that management needs to be among the change leaders.

2.2.4 Management status

Management status is not as objective as role and competencies because it is partly

determined by the opinion of team members. Looking at the theory on management status, two main constructs are identified. Mintzberg (1985) states that status is derived from

authority within the organization. Whereas Finkelstein (1992), defines power as the ability to apply your will, which is closely related to the status of a manager.

(15)

Authority

Authority was first defined by Simon Herbert (1951), as the right to choose actions that affect part or the whole of the organization. Aghion & Tirole (1997), and divided this authority into formal and real authority. Where formal authority is given due to an implicit or explicit contract which gives the right to make decisions, real authority is the effective control of decisions (Aghion & Tirole, 1997). For example, the president of a country does not really control all decisions made but only a small number of the executive branch. Theory indicates that real authority is distributed to subordinates if the principal does not have enough time to gather all relevant information to make an informed decision (Graham et al., 2015). In relation to this, Aghion & Tirole (1997) also indicate that a CEO is less likely to delegate authority when he/she is very knowledgeable and therefore has a lower cost to reach an informed decision. Decisions of capital allocation (internal investments) are more the responsibility of employees, as they are closest to the decision (plant/division level and less valuable) (Aghion & Tirole, 1997; Graham et al., 2015). Expending to a new product line or industry is more related to the global perspective/information and therefore the decision authority of the CEO (Harris & Raviv, 2005). CEOs that care more about a decision (mostly when pay is incentives-based) are less likely to delegate decision making control (Aghion &

Tirole, 1997; Hart & Holmstrom, 2010). As CEOs rely increasingly on data instead of the cognitive ability of the manager, the authority of the decision is likely to also shift to data analysts and scientists (Provost & Fawcett, 2013).

Power

Management has always been involved in determining strategic directions. Power is defined as the capacity of someone to apply their will (Finkelstein, 1992) and is therefore closely related to authority. Finkelstein (1992), proves the importance of four main power constructs:

structural, ownership, prestige, and expert power. Structural power is based on formal

(16)

positions, ownership power considers shareholdings, prestige power is driven by board roles together with elite educations and expert power is determined by the ability to handle

environmental dependencies (Finkelstein, 1992). Looking at the influence, Tushman (1977), argues that managements’ power has most influence with high level or non-programmable decisions, meaning that managers with a high level of power determine the course an organization takes (Finkelstein, 1992). Managements’ intuition, based on expertise and previous experiences, is leading in these business decisions and often results in effective outcomes (Burke & Miller, 1999; Salas et al., 2010). The intuition or deductive reasoning about decision making does however imply bounded rationality (Puranam et al., 2015).

Humans have limitations with information processing, meaning that not all alternatives in the process of decision making are considered (Feldman & March, 1981). Although this problem could be mitigated by delegating decision-making authority across roles with

interdependencies (Aoki, 1986), managers with higher power still have more influence in this process. In this context, Akhtar et al. (2019) proved that Big Data Driven (BBD) teams and actions are directly correlated with higher business performance. The way business decisions are made is changing and data and technology are gaining importance (von Krogh, 2018).

Looking at the power constructs of Finkelstein (1992), expert power will be under pressure, as the intellectual and cognitive proportions of technology have no boundary and will, most likely, in time outperform the cognitive capabilities of humans (Frey & Osborne, 2017;

Haenlein & Kaplan, 2019). To understand the management perspective on sharing of leadership with a synthetic team member, the current changes in the insurance industry context are important to consider.

2.3 AI in the insurance industry

As elaborated in the introduction, the insurance industry is selected for this research as AI in the sector has been increasingly implemented in processes such as fraud detection, claims

(17)

processing and customer targeting during the last five years (Soberanis, 2021). Although it is considered as an old-fashioned or traditional industry, the insurance industry is starting to exploit the exposure to large amounts of data. As AI can compress complex and large amounts of data into manageable scale, the opportunities are increasing in all departments.

Balasubramanian et al. (2021) even predict that the insurance value chain will experience drastic changes by 2030, like distributing insurance instantly, full automation of

underwriting, and claims with the Internet of Things sensors directly connected to the insurer.

Organizations will need to adjust and create a strategy that is more technology driven (Adner et al., 2019). This strategy should not just be pushed by the IT department but the entire organization needs to understand and commit to the changes ahead (Balasubramanian et al., 2021; Rader, 2019). The industry that used to be static and mainly driven by managements’

power will evolve continuously in the coming years and the need to be technologically adept will become more important (Aboelmaged, 2014; Finkelstein, 1992).

Because of the rapid developments and increased impact of AI in the insurance

industry, the Dutch Authority for Financial Markets (AFM, 2020) developed a report with the subsequent risks involved. This was also driven due to an AI driven mistake in childcare benefits, resulting in a country wide scandal (Schuurmans, 2021). The report of AFM advises on future laws, with an increased importance of policy, governance and explain ability of AI (AFM, 2020). The legal as well as ethical boundaries of AI are crucial to keep in mind for future innovations. A mistake in the design of AI could have severe consequences to the organization’s image, as seen with the scandal (Schuurmans, 2021).

To conclude the literature review, management is changing as AI is having an

increasing impact. The literature introduces several angles on the effect of AI implementation on the competencies, roles and status of management, however most are subtracted from parts based on different relations. There is still little research on the relation between changes

(18)

in management due to AI implementations, and therefore this research started with the aim of developing theory and conceptual frameworks.

(19)

3 Research Design and Methodology

3.1 Research Design

The research has explored the following RQ; how are the role, competencies and status of management affected by AI implementation? The literature review indicated that AI implementation has influence on several levels of an organization but also showed that previous research has limited information on the relation between AI implementation and changes in the role, competencies, and status of management. This gap drives the need for an inductive, explorative approach (Saunders et al., 2019). Whereas studies with an abductive of deductive character build upon theory, this research has limited theories to use (Saunders et al., 2019). The research, therefore, has a grounded theory approach where qualitative data gathering aims to explore new concepts and phenomena and to theorize their relation (Chun Tie et al., 2019). In line with this, the insurance industry was selected because of the increase in AI implementations in the sector over recent years and the impact of both these AI

implementations on the organization and on society. Interviews were conducted with

individuals to reach an understanding of this social phenomena. As opposed to focus groups, this gives a broader overview without the other opinions influencing the informants point of view.

3.2 Research Sampling

A form purposeful sampling was used (Trochim et al., 2016), selecting only management functions, within the insurance industry, and were required to be experienced both in the period before and after AI implementation. This was a must for the data collected as the differences between the two is the core of this research. The demographic area will be focused on the Netherlands and Belgium.

(20)

3.3 Sample Participants

A total of eight informants have been interviewed. With the help of my personal network, I have contacted a large part of the population directly. The participants used have all been anonymized. Results of the research in the form of an executive summary will be shared with the respondents afterwards as this was used as incentive. An overview of the informants can be found in Table 1.

Job title Company/Team size Experience Head office Inf. 1 Director IT and Operations 700 FTE 15 Years The Netherlands

Inf. 2 CEO 100 FTE 42 Years The Netherlands

Inf. 3 Department head non-life insurance 180 FTE 12 Years The Netherlands

Inf. 4 Partnerships Manager 350 FTE 30 Years Belgium

Inf. 5 CEO 1200 FTE 4 Years Belgium

Inf. 6 Board member and manager of injury claims

200 FTE 9 Years The Netherlands

Inf. 7 Department head Digital Marketing 100 FTE 4 Years The Netherlands Inf. 8 Department head Analytics 80 FTE 12 Years The Netherlands

Table 1: Profile of Informants 3.4 Data Collection

With the use of semi-structured interviews through the application Microsoft Teams, this research was able to contribute to the exploration of the specific topic. In contrast with structured or unstructured interviews, it applies structure while maintaining the flexibility of exploring the constructs in-depth (Fossey et al., 2002). The goal of this research was to reach data saturation level, and this was expected to be reached with six to twelve interviews (Guest et al., 2006). Saunders et al. (2019), define data saturation as the point where new data

repeats that which was said in previous data. This means no new insights on the effect of AI implementation on the role, competencies, and status of management.

(21)

To reach informed consent and promise anonymity, a form of consent was created and signed by the informants as well as the interviewee (see appendix 2). In order to let the

informants get acquainted with the concepts and purpose of the study, an introduction letter was distributed beforehand (see appendix 3). The script of the interviews was also prepared beforehand and can be found in appendix 4. The interviews took between 30 to 45 minutes and some interview questions changed slightly as data was gathered and different insights were discovered.

3.5 Data Analysis

After the transcription of the interviews, the qualitative data needed to be analyzed. This was done through the software of NVivo, which is specifically designed for dealing with this process (NVivo, 2022). While the research aims to discover and generate new concepts, it remains important to achieve "qualitative rigor" (Gehman et al., 2018).

To do so, this research follows the Gioia methodology (Gioia et al., 2013). Although the differences are minimal in qualitative research methods, Gioia is considered as a method that reaches maximum rigor with inductive research (Gehman et al., 2018). As there is not much theory in the literature, the interviews are approached with less predetermined ideas and a "willing suspension of belief" (Gioia et al., 2013), to avoid a conformation bias. With the use of NVivo, the eight interviews were approached with this open mindset. The method starts with quotations, which in turn led to the 1st-order categories (open coding), which is a broad set of codes that are seen as a descriptive view of the findings (see appendix 5).

Afterwards a helicopter view to all 1^st-order categories and known theoretical concepts, decreased the amount of categories and this, in turn, led to the creation of 2nd order themes (see appendix 6). The emerging themes were then used in the explanation of the phenomena of the research and aggregated into overarching dimensions (see appendix 7). All three orders of analysis were then combined in the data structure to visualize the findings within the data.

(22)

The data structure created the basis of the research and the grounded theory. In this it is important to, not only understand the concepts, themes, and dimensions, but especially their interrelatedness.

Although this is qualitative inductive research, the Gioia approach applies rigor through its systematic approach. The assumption that the informants are knowledgeable agents, is however crucial as this emphasizes the importance of their knowledge to identify new concepts and relationships that are outside of our previous understanding (Gioia et al., 2013). Furthermore, the systematic analysis of the data provides a more evidence-based presentation to counterbalance the assumption that I can trace and surface constructs and relationships that might go unnoticed by the informants. Furthermore, dependability is reached as transcripts and audio tapes of all interviews are available upon request.

Looking at the informants used, it is a hard to access a group of highly knowledgeable people of the concept and relation targeted, which makes the use of the researchers network a strength of this research. That being said, the minimal experience of the researcher in

conducting academic research, and especially in conducting interviews, is a limitation that should be considered.

(23)

4 Findings

The findings from the interviews are displayed in this section and will highlight the most important changes in the role, competencies, and status of management after AI

implementation. The impact of AI on the insurance industry is also elaborated upon and although this is not directly related to the RQ, it does give more context to the changes within the sample of managers due to AI implementation. The interviews analysis yielded 35 (11 for context and 24 to answer RQ) categories which the researcher organized in 12 themes and 4 dimensions. All dimensions are elaborated upon separately in the findings after which the discussion is focused on the three dimensions that consider the RQ. This showcases how it supports current literature and contributes to fields not yet defined in current research or theory.

4.1 Impact of AI implementation on the insurance industry

The literature gave an indication of the changes within the insurance industry, and here the research highlights the impact AI has on organizations from the perspective of the

informants. The findings are supported with the data structure of the impact of AI

implementation on the insurance industry dimension. In table 2 a part of the data structure is visualized, for the full structure please refer to appendix 8.

4.1.1 AI implementations in the insurance industry

The insurance industry used to have many standardized manual processes and paperwork.

Management needed to motivate employees to do a lot of “machine” work to make sure the stacks of claims were being processed through. The function had a need for a people manager with a specific domain knowledge to assist on difficult cases, as inf. 1 states: “In our

company you no longer have team leaders who manage stacks and processes and make sure employees stay motivated to do this.”. Nothing was furthermore done with the processed claims, and they were just archived (Inf. 1). All managers therefore interpreted a situation

(24)

individually and this caused variance as Inf 6 indicates with: “Everyone worked quite individually and that resulted in variance in what we had to reserve”.

The introduction of AI was however also not the immediate solution. The functions were not contributing to the process as humans were able to do the same task in the same or even less time (Inf. 5). This was the first impression of AI and it may have caused some doubt about the possible added value.

The implementation of AI is now rolled out in several departments, like claims handling, sanction screening, marketing, and fraud detection. From automated processes like the direct payment of simpler claims (Inf. 3) or reading document and extracting the right information (Inf. 5). These implementations have developed previous human powered processes of long phone calls and paper trails to more efficient and AI powered processes, as Inf. 4 elaborates: “Today we manage to handle an accident with own damage fully

automatically in 60-70% of the cases”. Processes are automated or human processes are augmented with AI, in almost all areas of an insurance firm, with less employees, improved service towards the customers and improved scalability. This affects the firm’s business results as Inf. 1 states: “The profitability of our services has increased enormously in the past 5 to 10 years and the margins have really at least doubled.”. This is achieved as processes get automated but also as the organization can look ahead, moving from analysis to

prediction with data (Inf. 8). With the increased impact on the results, the managers are incorporating the developments from a strategic point of view as Inf. 5 elaborates that they need to embrace and intergrade the strategic component (AI) into their strategy.

The AI implementations also assist the customer with pre-filled information giving them more time to focus on important questions and enables them to report a claim 24/7 (Inf.

1 and 2). With the improved business results and customer service, the developments in the industry are rapid and Inf. 5 thinks that if you fail to do so with your own organization, you

(25)

will become irrelevant. However, Inf. 7 also states that: “In particular, to properly convey that emotional side to the customer, a bit of empathy and understanding, huh, that's digital, is that still very difficult”, indicating that humans are still very important in certain elements of the customer experience.

Although most of the informants were already convinced about the positive impact of AI implementation on business results, Inf. 3 believes that AI is still underdeveloped to truly impact the organization, stating: “That (AI causing job replacement) is still so far in the future that it isn't threatening, so to speak, so that (our attention to AI) was just a healthy enthusiasm and about innovation”. This is important to mention, as it indicates that not all managers are fully convinced and in support of AI adoption.

1st-order Categories 2nd order Themes

A I i m pac t on t he ins ur anc e i ndus tr y

Old way of insurance

Before AI implementation Claims handling before AI

AI used to be

AI implementation for predictions

AI proceses in financial services AI implementation in Claims

AI implementation in marketing (e.g. Churn models) AI indicating fraud process

AI Automated document processing/analysis AI improving business results

AI driving business results Improved customer experience due to AI

AI is still underdeveloped

Table 2: AI impact on the insurance industry

(26)

4.2 Management Competencies

This part explores the impact of AI implementation on the management competencies needed. The findings of the interviews indicate that management needs to improve

competence in change management and that AI knowledge is crucial for other competencies.

This is supported with the data structure of the management competencies dimension: in table 3 a part of the data structure is visualized, for the full structure please refer to appendix 9.

4.2.1 Change management

When the informants look at the AI implementation impact from a broader perspective, they indicate that it requires change management capabilities, with Inf. 4 stating: “I think that most of the competencies that you develop, where you, from my position, that I have developed, are competencies around change management”. To support management in this transition, change management tools are useful to make sure that all employees are aligned. However, the starting point for the informants is carrying out the vision of the change, as inf. 4

indicates: “You must have a very clear vision from management to ensure that that everyone is working towards the same direction”.

4.2.2 Developing AI knowledge

As stated in “Impact of AI implementation on the insurance industry”, AI is becoming a larger part of the organization and in line with this management is required to develop knowledge on the matter. All informants indicate to have invested time into learning the basics of AI. Inf. 7 elaborates on it with: “But what it asks of me as a manager, I need to understand it, I don't have to be able to do it myself. I don't need to be able to code it myself or literally know how the code is written. But I need to understand the logic, the methodology behind it”.

This also helps as management is approached by several AI/fintech developers. As AI is making ground in several industries, specialized services are offered, like inf. 5 states:

(27)

“Our competence is not in coding the algorithms behind the AI techniques. No, there are companies that are fully focused on that, but we want to act through partnerships with these companies.”. The decision to outsource that often lies with management, again indicates the need to develop AI knowledge.

The informants further elaborate that understanding AI basics, helps in the decision to use it with automation or for augmentation of human processes. Inf. 1 and 6 indicate that deciding on different solutions depending on the process is an important part of management competencies. As developments continue the managers try to move from augmentation to automation, with Inf. 4 stating: “In the past we were making tasks of employees easier and today we try to take those tasks, I'm saying it disrespectfully, but take it away, and let the computer take over”. As progress moves on, Inf. 2 points out the importance of version control in assessing future steps of the algorithm.

The source of intelligence is data, indicating the importance to understand and to think from a data point of view. The cloud developments have enabled organizations to gather more data and make it available for AI implementations (Inf. 2). It is important to understand the data need of AI, as Inf. 1 indicates: “I'm old school you might say, so I've really had to delve into data knowledge, as we have a lot of meetings on data opportunities.”.

With AI functionalities improving, the need for data might change, pushing organizations to structure, the currently “useless” data, in a way that it is easily accessible for possible future AI implementations (Inf. 4). With the increased importance of data, inf. 7, 8 and 2 indicate that it is important to also get a grip on the ownership of the data and the algorithm, as well as the basics on how the algorithm uses the data. Furthermore, as AI is incorporated in the process, personal knowledge becomes less important and moves more towards the usage of data, or in other words generalized knowledge (Inf. 1 and 4). As Inf. 6 states: “I can have in

(28)

my head what I know myself, so my own experiences. And what that model does is add up my experiences and the experiences of all colleagues.”

The generalized knowledge also changes the approach towards solutions.

Management needs to use resources and insights differently, as Inf. 2 states: “if someone says to me, I want to make processes STP, for example. And that's 10 steps and they want to digitize it, while I know that, by connecting an external data source, we can make 6 steps superfluous”. Inf. 8, elaborates on process management with better insights as he reviews the entire process once every 3 weeks. To improve the process the manager sometimes even needs to look at the needs and capabilities of AI as part of the team, as elaborated by Inf. 2:

“Once that algorithm is in place or once the digital process is in place, then it is very easy to add the following external data source, for example. Or to fine-tune the algorithm a bit or to add an extra algorithm and that is really a different way of thinking, so to speak.”.

Expectation of technological advancements within departments is rising and management is required to act upon it (Inf. 1).

After successful AI implementation the scalability increases, as Inf. 1 states: “if we, for instance, have to make quotations, we can now do 1000 in a day where you used to be able to do 10.”. With scalability increasing the impact, a single mistake could affect 5 million customers instead of 30 (Inf. 2). The entire organization needs to understand that actions not only have greater impact, but also enables opportunities, as many operational managers can look for AI assistance instead of adding more employees as soon as the workload increases (Inf. 2).

The insurance industry is closely monitored, and rightly so is the opinion of Inf. 2, as the increased amount of personal data involved in the day-to-day business needs to have a higher level of data governance. Everyone in the organization is trained to increase the level of knowledge with regards to the GDPR regulations, as this was confirmed by all informants

(29)

in this research. Some of the organizations even test all proposed action involving data on GDPR and compliance (Inf. 4). Customers are also increasingly concerned about their data and Inf. 8 elaborates on the importance for the long-term with: “We want that in 10 years' time, customers say to us, it is great to be a customer with you, because you have always handled our belongings with care.”.

Ethics is another critical point to consider for the implementation of AI, with risks of skewed data causing biases and in turn discrimination risks. As Inf. 3 highlights the

importance using the example of a recent Dutch scandal: “Because as a brand you don't want to be associated with things like that, especially if you look at what happened with the

childcare benefit scandal, yes, they might be labeled, for systematic racism”. However, if competition does implement certain algorithms, as they might interpret rules slightly more lenient, you might be pushed to do so as well (inf. 2 and 7). Being a manager in these conditions, you need to balance regulations because you do not want to “slam everything shut” (Inf. 7), but you also not want your employees to push for the edges every time.

M ana gem en t C om pet en ci es

Emphasis on change management Increased importance of change management Vision important

AI knowledge needed

Developing AI knowledge Dealing with AI scalability

Automation & Augmentation of AI

AI changes expectations and the way of thinking Data based targeting

Skill to look at business from a data point of view Data based thinking

Management less on personal knowledge but on how to use generalized knowledge

Discrimination risk of AI

Ethics and privacy risks of AI implementation Ethics standards for insurance

Ethics standards for insurance GDPR knowledge needed

Table 3: Management Competencies

(30)

4.3 Leadership Role

The leadership role is also changing due to AI as the organization becomes more data-driven and employees fear possible job replacement. The findings of the interviews are set out here and supported with the data structure of the leadership role dimension. In table 4 a part of the data structure is visualized, for the full structure please refer to appendix 10.

4.3.1 Changes in leadership role

The consensus of all informants is that the needs of their teams are changing. With the industry moving from human driven to more data driven, Inf. 1 elaborated: “it's more of a technical analysis of the company or business unit than a substantive discussion about how people are doing, that's also part of it, but it's more data-driven than people-driven”, and it also explains why the respondents are seeing their role existing more in between Business and IT. It is moving to process management instead of people management. Where Inf. 2 states “the organization no longer has managers responsible for ensuring that his/her employees, for example, handling the large piles of claims, stay motivated”. The leadership role is also focused on making sure everyone understands the entire process as Inf. 4 states:

“employees who fail to think in a process-oriented way, fall by the wayside, even as an operational employee. Because today you need to know a lot more about the process as a whole to be able to fill in the part of the process”. AI therefor also changes the kind of team you manage, as simpler functions are being automated, the team consists out of heavier profiles (Inf. 4).

There is a shortage in the general labor market, but digital talent is even more hard to find (Inf. 2, 7 and 8). Inf. 2 is even expanding the recruitment of digitally skilled employees to other countries like Spain and Poland. To attract and retain new talent, organizations are offering digital training programs and training courses to employees, but management needs

(31)

to enable the team to take on exciting AI projects. As Inf. 1 and 2 state that people who work with it find it “cool” or say they “love it", indicating the need to encourage such projects.

The implementation of AI also means that certain jobs become redundant. This causes stress and uncertainty as Inf. 3 states that employees are saying: “what are you doing here?

Am I going to lose my job?”. This emphasizes the role of the manager to communicate transparently and frequently about the changes ahead. 75% of the informants indicate the importance of pushing for employee development and training. This is in line with explaining to employees that they will be enabled to do more valuable and relevant work due to AI implementation (Inf. 1 and 2). Resistance towards change is a logical first human reaction but helping employees thrive in this new era, leads to employee retention (Inf. 3 and 4).

Competition for digital talent is fierce and the role of management will need to adjust to suit the new needs after AI implementation.

lea der sh ip R ol e

Different employees in team

New kind of team needs From simple functions to heavier functions

Managing team needs of IT and Business is getting more important

Digital skilled employees are hard to find Attracting and retaining digital talent Digital talent & traineeship

Employee resistance

Dealing with resistance Explaining change to employees

Focus on internal employee development

Table 4: Leadership role

(32)

4.4 Management Status

The role and competencies are changing as AI is implemented, making it more focused on the process and is placed between business and IT. This also causes the need for different

structures and teams as well as a hierarchy shift. The findings of the interviews are set out here and are supported with the data structure of the management status dimension. In table 5 a part of the data structure is visualized, for the full structure please refer to appendix 11.

4.4.1 Changes in management status

The hierarchy has been changing due to digitalization and AI is also a driver in this as Inf. 7 explains that a layer in the organizational structure was removed and Inf. 1 elaborates that where it used to be the profession knowledge which had most to say in the organization, this is now more towards the ones who are experts with data and processes. Not sure if this is already the case, but as elaborated in the competencies part, the usage of personal domain knowledge becomes less important, meaning that the expert power is less meaningful and moving more towards data analysts and scientists (Inf. 4).

With departments becoming more dependent on each other after AI implementation, 50% of the informants indicate that their organization is adopting an agile way of working.

The main reason for the shift is highlighted by the need to have teams of both business and IT people (Inf. 2 and 7). This allows the teams to make the translation between the two easier and gives more flexibility to the projects executed (Inf. 4). It means that business and IT are merging more and more, as elaborated in the management competencies dimension that the insurance firms are transforming into IT firms, as Inf. 7 states: “our teams are also complete IT and business roles mixed together, so we no longer see it separately”. The authority of digitally skilled employees is increasing, as they need to take more departments to the next level, as a department head (Inf. 6) explains: “we let ourselves be coached a lot and it works very well”.

(33)

It is also shown in the knowledge that is needed, as specific insurance knowledge is moving more towards the background, elaborated by Inf. 1: “That means that certain skills such as specialist knowledge are more in the background for larger claims that do not fit into the system”. Where an MT meeting used to be with no data analysists at all, now there are four (Inf. 1). As indicated in management competencies, a manager’s specific knowledge is less leading, this also affects their status within the organization. The informants in some form agree that the insurance industry and its managers are changing due to AI

implementation, but it is not yet that clear from a management point of view.

St at us M ana gem en t

Flat organizational structure

Hierarchy changes More power towards data analysts

Agile to support AI transformation

New departmental structures Training other departments in AI

Table 5: Management Status

(34)

5 Discussion

In this part the findings are used to answer the RQ, and compared to previous research, provide the base for recommendations on future research.

5.1 AI implementation changing management competencies, role, and status

This study will add to the research enterprise with a contribution to the growing field of research on the impact of AI (Danyluk & Buck, 2019; Raisch & Krakowski, 2021; von Krogh, 2018). AI implementations have increased in almost all industries and the need for better academic understanding is needed (Haefner et al., 2021). The increase is especially shown in the insurance industry, hence the selection of this specific domain. Within AI research most attention was distributed to the technological and organizational implications, and less so on management (Mikalef & Gupta, 2021; Raisch & Krakowski, 2021). This research goal was to investigate the effect of AI implementation on the competencies, role, and status of management. Therefore, the study aims to answer the following RQ: How does AI implementation affect the competencies, role, and status of management?

The findings suggest that there are indeed changes in management due to AI implantation. As research on AI capabilities suggested, the sole deployment of AI is ineffective and to stay competitive and adopt AI in the organization, among other things, different management skills need to be developed (Conboy, Mikalef, et al., 2020; Mikalef &

Gupta, 2021). As AI has been implemented in the insurance industry, the changes in

management become more apparent with, for instance, the need for AI competencies, dealing with employee resistance and new departmental structures.

5.1.1 Changes in management competencies

The informants pointed out several competencies that were needed or that changed due to AI implementation. An overview of the different competencies can be found in figure 1.

(35)

The first competence is the need for change management capabilities as a manager.

This is in line with what Mikalef & Gupta (2021) indicate, that business value through AI investments is only achieved when leaders understand and commit to drive a large-scale change. This approach to AI implementation required the informants to embrace a change vision and carry it out throughout the entire organization. Although this is closely related to the leadership function, the skills and attitude in organizational change management is now considered almost a necessity in order to succeed in the competition fueled business

environments (By, 2005; Hornstein, 2015). The need for managing change is also in line with disruptive innovation theory, with one of the enablers being: “Prepare for and instituting organizational change and unlearn its deeply entrenched values” (Yu Dan & Hang Chang Chieh, 2008, p. 407).

The second competence the informants identified is the development of AI

knowledge, from which multiple knowledge needs are formed. Theory on management skills in AI capabilities also hinted in this direction with the need for managers to be able to

identify possible opportunities with AI in their domain, as well as managing the transition to AI powered processes (Mikalef & Gupta, 2021). Referring to the definition of Moore et al.

(2002), in possessing the required skills, knowledge and attitudes, this knowledge is key to develop the right competencies. With all informants investing time into understanding AI, it indicates that it is of great importance. Kiron (2017), confirms and foresees that the manager needs to develop new skills and knowledge. The first derived competence is deciding if an AI implementation needs to augment humans in their process or, if it is possible, automate the process. Furthermore, in this decision it is important to understand if an algorithm can improve with data gathering and move towards automation, as elaborated upon by Raisch &

Krakowski (2021).

(36)

The informants also indicate that AI makes personal domain knowledge less

important. With AI’s current capacity to combine the knowledge of the entire organization in its decision making, the impact on different knowledge needed will increase as more data is made available for AI (Mitchell & Brynjolfsson, 2017). The role of the manager changes from being focused on exploiting personal knowledge to where and how to use the combined knowledge of the organization (Rhem, 2021). As a recent study reveals a need for other skills: “The managers we surveyed recognized the value of judgment work. But they undervalued the deep social skills critical to networking, coaching, and collaborating that will help them stand out in a world where AI carries out many of the administrative and analytical tasks they perform today.” (Kolbjørnsrud et al., 2016, p. 5). This will, in addition to the development of new skills, also influence the power of management which is discussed in the changes in management status.

After AI implementation the scalability increases, resulting in better business results with the margins doubling (stated in the findings of “impact of AI implementation in the insurance industry”). It, however, also increases the impact of mistakes, which is something the manager and the team should be fully aware of. A mistake in the system could lead to harmful behavior, potentially towards large groups of customers due to scalability (Amodei et al., 2016). This also links to the ethical and privacy risks of the AI and the accompanying data. Insurance is a risk averse industry as it is the core of their right to exist, meaning that these risks are receiving much attention across all informants. The competence to manage Ethics and privacy risks of AI needs a level of AI knowledge as well as clear communication of the organizations business ethics (Snoeyenbos et al., 2001). The possible implications of mismanagement became painfully clear in the Dutch childcare benefit scandal, with the risk of being labelled as an organization that used systematic discrimination (Schuurmans, 2021).

(37)

The third competence is the ability to think from a data point of view. This entails that the manager should see future opportunities in the available data and act upon it as AI is only useful if it can interpret and learn from data, like it was stated in the definition of Haenlein &

Kaplan (2019). Looking at the theory from Mikalef & Gupta (2021), the domain knowledge could be combined with the ability to think from a data point of view opening up more opportunities for the future.

Figure 1: Changes in management competencies 5.1.2 Changes in leadership role

Due to AI implementation, the informants experienced several changes in their role. An overview of the changes can be found in figure 2.

Starting with a broad change in the role, as the findings indicate that AI drives different team needs. With the MT meetings containing more technical discussions,

management is required to adjust their role and make sure the needs of the team are secured, with AI as a team member taken into account (Morgeson et al., 2010). The aim of a team leader is to enhance team effectiveness by meeting the core team needs. This is achieved by affective emergent states, cognitive emergent states and behavioral integration processes (Morgeson et al., 2010). The findings indicate that the business is becoming more data driven and process oriented, which puts the ability of a team member to identify with the

organization under pressure (affective emergent states). Furthermore, the team adjustment to

AI implementation

Increased importance of change management

AI knowledge development

Ethics and Privacy risks of AI implementations

Skill to look at business from a data

point of view

(38)

the integration of AI as a team member is altering the cognitive emergent states. This suggests that the leadership role needs to adapt to reach team effectiveness.

With informants point out the importance to think in a process oriented way, the role of leadership is to make sure their teams understands the need for knowledge and to

understand the entire process. This is in line with the transformational leader theory, where management encourages the team to be committed to the organization’s current activities (Dvir et al., 2002). Avolio et al. (1999) elaborate that the role of a transformational leader consists out of four elements, and highlight one as intellectual stimulation, which achieves the involvement of the team in the entire process. This indicates that management should encourage employees to think beyond their traditional boundaries.

With the managers and employees more focused on how and where to use the right resources and insights, the possibilities for innovation increases (Ågerfalk, 2020). As

innovation and AI implementations play larger roles in the organization, the operational team realizes that their job might become redundant, which causes resistance that needs to be managed. The management of the increased dominance of the technological side and the fear of job loss on the operational side of the organization, is now part of the leadership role (Kiron, 2017; Kolbjørnsrud et al., 2016). Employee training and development is indicated as a focus point for management as 75% of the informants say it is crucial in managing

resistance. This is reflected in literature where Fountaine et al. (2019), indicates the

importance of transparency and development to counter employee resistance. Looking at the four elements of transformational leadership, individualized consideration could be highly effective as it focusses on developing the potential of employees (Avolio et al., 1999).

The importance of retaining these employees was only increased over the last year as the labor market is very tight (Duval et al., 2022). Training employees internally to fill different kind of functions could work, however the complexity of the vacancies increases as

Management: Adjusting to the Newest Artificial Team Member